Introduction
Advances in medical treatment have significantly improved the prognosis of human epidermal growth factor 2 (HER2)-positive early breast cancer (BC) patients over time and led to establish chemotherapy combined with 1 year of trastuzumab as the standard adjuvant treatment [
1].
The impact of prognostic/predictive biomarkers on the outcome of patients treated with appropriate standard systemic treatment has been considered by the American Joint Committee on Cancer (AJCC) Staging System panel in the update of the breast cancer staging. Based on the incorporation of biologic factors (histologic grade, estrogen receptor, progesterone receptor, HER2, and multigene panels) to the classic anatomic stage, the 8th edition of the AJCC breast cancer staging system has introduced prognostic stage, which was developed using data from patients identified in the National Cancer Database (2010–2011) and then validated in large cohorts of patients from the MD Anderson Cancer Center and the California Cancer Registry [
2‐
7]. These studies allowed to confirm the improved prognostic performance of the prognostic stage as compared to the anatomic stage in the general breast cancer patients’ population. The most recently updated version of the prognostic stage was released after the results of the validation study highlighted that a proportion of patients could not be assigned a specific prognostic stage [
7]. Therefore, the prognostic staging system was refined to include all the possible combinations of anatomic stages and biomarkers [
8]. As declared by the AJCC staging panel, the actual prognostic stage will undergo frequent updates, based on future validation studies in large databases of patients treated with state-of-the-art therapies [
4,
6]. Several studies, all conducted in retrospective patient cohorts, have been reported in the last couple of years, overall corroborating the prognostic stage as a more accurate discriminator of breast cancer patients’ outcome as compared to the anatomic stage. However, it has to be pointed out that many of these studies used data from the National Cancer Database or the SEER (Surveillance, Epidemiology, and End Results) registry covering a period of time including years 2010 and 2011. Considering the overlap between the National Cancer Database and the SEER, these studies included data that were previously used by the AJCC panel to develop the prognostic score. Moreover, most of these studies, including the main validation studies by the AJCC panel, did not report detailed analysis of distinct breast cancer subtypes, with no study specifically focused on HER2-positive disease. Furthermore, even in the most robust cohorts, exposure to trastuzumab was not reported or not homogeneous among HER2-positive patients (literature review in Additional file
1) [
7,
9‐
23]. This aspect is a relevant caveat, since the assumption at the basis of the adoption of the prognostic stage is that patients are offered adequate systemic treatment based on biologic characterization [
2,
4,
6].
One of the major clinical needs for HER2-postive BC patients is an accurate risk stratification to guide escalated and de-escalated strategies to ensure the most effective treatment along with a more rationale resource allocation [
24]. One of the most important goals of staging is to help clinicians define a treatment plan [
5]; therefore, the evaluation of outcome prediction by the prognostic stage in HER2-positive patients cohorts treated with standard therapy is a key step in order to define its potential role as tool to guide de-escalated therapeutic choices. For this kind of investigation, a randomized trial testing de-escalated against standard treatment represents the ideal setting.
In this study, we aimed to validate the prognostic stage in HER2-positive BC patients treated with adjuvant chemotherapy combined with 1 year or 9 weeks trastuzumab in the randomized ShortHER trial [
25].
Discussion
This is the first study (i) evaluating the performance of prognostic AJCC stage specifically for early HER2-positive BC patients treated with adjuvant chemotherapy and trastuzumab, (ii) evaluating the performance of prognostic AJCC in a prospective randomized trial, and (iii) validating the prognostic AJCC in a European patients’ cohort. Our findings show a similar prognostic performance for prognostic and anatomic stage, despite prognostic stage reallocated a substantial proportion of patients (41.6%) to a more favorable stage category. Previous studies in general BC patient populations have described a reallocation rate with prognostic stage most frequently reported around 40–60% (range 18–74%; Additional file
1) [
7,
9‐
23]. In the California Cancer Registry, including 54,727 patients in anatomic stages I to IV, 31.0% and 20.6% of patients were assigned to a more favorable and less favorable stage category with prognostic stage, respectively [
7]. Only a few studies reported the discrepancy between the two stage models specifically for HER2-positive BC patients, leading to non-univocal results (Additional file
1) [
9,
14,
21,
23]. A large cohort from the National Cancer Database including
n = 60,155 HER2-positive BC patients showed 29.4% and 0% rates of downstaging and upstaging, respectively [
9]. Another study showed that 35.8% and 40.7% of HER2-positive patients (
n = 1982) were classified as stage I by anatomic stage and prognostic stage, respectively [
15]. The rate of downstaging (58.4%) was higher in our study, and consequently, the enrichment in stage I patients with prognostic stage was also more evident. The high prevalence of hormone receptor-positive patients in the ShortHER population (68%) might have contributed to substantial downstaging. It should be highlighted that the ShortHER population reflects the characteristics of HER2-positive patients commonly treated in contemporary clinical practice [
25,
27].
Our data show that the substantial downstaging of patients with the prognostic stage did not affect the performance of the model which was maintained similar to anatomic stage (
P = 0.975 for C index comparison). In this context, available literature data focused on HER2-positive patients are scanty. Moreover, their interpretation is extremely limited by the lack of homogenous treatment or lack of information about it (Additional file
1) [
18,
23]. The largest cohort of HER2-positive patients analyzed for survival outcome according to prognostic stage included 562 cases (mostly not treated with trastuzumab) and showed a good 10-year disease-specific survival for prognostic stage I patients (> 96%), but did not report overall model performance [
18].
As previously discussed, prognostic stage led to an enrichment in stage I (70% vs 40.9% anatomic stage) and more specifically in stage IA patients (58.9% vs 37.7%). The pairwise comparisons conducted in stage I–IIA patients suggest that the prognostic stage better discriminated the group of patients with the best prognosis among others (IA), whereas with anatomic stage there was no significant difference in outcome among patients in stages IA, IB, and IIA. However, when looking at absolute survival rates, stage IA patients had slightly numerically inferior outcome as compared to stage IA groups by anatomical stage. The numerically worse outcome for prognostic stage and matched anatomic stage categories was evident for all stage groups. In synthesis, the prognostic stage, by recognizing the prognostic effect of biomarkers, results in a shift from a worse to a better stage category (mostly to stage IA or stage IB) of a large number of patients as compared to the anatomic stage. One of the consequences of this shift is a better separation of stage groups in terms of DDFS, especially in stage I–IIA patients. However, intuitively, in absolute terms, the outcome of prognostic stage IA and IB patients, being enriched by patients with a worse anatomic stage category, is somehow diluted and results numerically inferior to the corresponding anatomic stage. Moreover, prognostic stage > IIA categories are depleted vs the same anatomic stage category in patients with a better prognosis; again, as a consequence, the outcome of prognostic stage > IIA groups is numerically inferior to the corresponding anatomic stage. The main implication is that the prognostic stage is more valuable as anatomic stage as a tool to counsel patients about their prognosis: by applying the prognostic stage, more patients would be regrouped in more favorable stage categories and would be informed about a good outcome as compared to the anatomic stage. To the other side, the prognostic stage identifies a more restricted number of patients with far poorer outcomes. However, what clinicians have to keep in mind when counseling patients is that in absolute terms the estimated outcome for a given prognostic stage category might not correspond to the estimation for the same anatomic category.
An appropriate identification of patients at excellent outcome with standard adjuvant treatment is key to identifying those patients who may be offered de-escalated treatment strategies. Treatment de-escalation for HER2-positive patients with anthracycline-free regimens as the paclitaxel-trastuzumab schedule is already administered in clinical practice based on anatomic stage, mostly for patients with stage I disease [
28,
29]. We explored whether prognostic stage I may be of value in identifying patients for de-escalated therapies. Our results suggest that if anatomic stage I seems a good parameter to guide de-escalated therapeutic choices, this may not be the case for prognostic stage I. Indeed, prognostic stage I patients treated with short trastuzumab had an absolute 3% worse DDFS rate at 5 years as compared to patients enrolled in the long trastuzumab arm. When restricting the analysis to prognostic stage IA, there was still an absolute 1.5% difference in 5-year DDFS favoring the long arm. However, this result was not statistically significant and was based on a difference of just six events between the two arms. Although these were exploratory, unplanned, and unpowered analyses that should be interpreted with caution, the results can be considered as hypothesis-generating that require further testing in similar trials. To note, the acceptable absolute difference in outcome to consider a de-escalated treatment as safe is currently debated [
30]. If our results will be confirmed in further studies, the two staging systems will be recognized as providing divergent information in the context of patient selection for treatment de-escalation, possibly posing a challenge in the implementation of the prognostic stage in clinical practice.
Our study has strengths: this is the first study evaluating the prognostic performance of prognostic stage in a cohort of HER2-positive patients, all receiving chemotherapy and trastuzumab; patient population is derived from a prospective trial; 99% of patients had sufficient data for the present analysis; the study design allowed to explore short vs long trastuzumab in stage-defined groups.
Main limitations of this study include the choice of the survival endpoint (DDFS) which is different from the one used to develop and validate the prognostic stage (BC-specific survival) [
7]. In the ShortHER trial, actual median follow-up does not allow for a mature evaluation of BC-specific survival in this population of patients. Therefore, we opted to use DDFS as a surrogate of BC-specific survival considering the lethal nature of DDFS events. Another limitation is the reduced sample size in stage-defined groups, limiting the power of direct comparisons.
Competing interests
MVD reports personal fees from Genomic Health, personal fees from EliLilly, personal fees from Celgene, outside the submitted work. AAB reports grants to organize courses from Devicor-Mammotome, Roche and Biocartis, outside the submitted work. AF reports personal fees from Roche, personal fees from Novartis, personal fees from Pfizer, outside the submitted work. AM reports grants, personal fees and non-financial support from Roche, personal fees and non-financial support from Lilly, personal fees from Pfizer, grants, personal fees and non-financial support from EISAI, personal fees from Macrogenics, grants from AstraZeneca, outside the submitted work. OG reports grants, personal fees, and non-financial support from Eisai, personal fees from Amgen, personal fees from Novartis, non-financial support from Celgene, outside the submitted work. CZ reports personal fees from Takeda, personal fees from Pierre Fabre, personal fees from TEVA, personal fees from Istituto Gentili, personal fees from Roche, personal fees from EISAI, personal fees from Novartis, personal fees from AstraZeneca, personal fees from Pfizer, personal fees from PharmaMar, personal fees from Celgene, personal fees from Lilly, personal fees from Amgen, personal fees from Tesaro, and personal fees from QuintilesIMS, outside the submitted work. FP reports personal fees from Eisai, outside the submitted work. MPF reports grants from Roche, grants from Devicor-Mammotome, personal fees from Biocartis, outside the submitted work, outside the submitted work. VG reports grants (Institution) and personal fees from Roche, personal fees from Novartis, personal fees from Eli Lilly, personal fees from Roche, outside the submitted work. PFC reports grants from Agenzia Italiana del Farmaco AIFA, during the conduct of the study; personal fees from Novartis, personal fees from EliLilly, personal fees from AstraZeneca, personal fees from Tesaro, personal fees from BMS, personal fees from Roche, grants (Institution) from Novartis, grants (Institution) from Roche, grants (Institution) from Merck KGa, grants (Institution) from BMS, outside the submitted work. All other authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.